The Smartphone CPU Review You NEED to Watch! on HackerTube: recommended by 4 Hacker News commenters across 3 separate months, 2022–2023.
explainer · intermediate · dated
Geekerwan reviews smartphone CPUs by power efficiency rather than benchmark scores, with test methodology and chapters on Qualcomm, Apple, Samsung and Google chips. It explains why high benchmark scores can coexist with heat and poor battery life.
HN people cite it as detailed data on mobile SoC efficiency, sustained performance and ARM core efficiency curves.
“explores the power efficiency graphs and that's probably what most phone users really want”
8 Hacker News comments link this video. Each one links back to its thread.
Interesting that geekerwan found the exact opposite in their testing of A710 ▶ this video ARM efficiencies at 13:10. Granted these are across multiple chips. The whole video is a good watch.
in “ARM’s Cortex A710: Winning by Default” · View on HN
Someone else posted a really compelling video where they test the power efficiency of Android and Apple phones. ▶ this video Its really quite informative.
in “ARM’s Cortex A53: Tiny but Important” · View on HN
This doesn't really showcase how much Apple's chips are as much as how far Qualcom has fallen behind. On the other hand, the results also show how much work Apple will need to put into their GPUs, as the clearly inferior chip is still beating the iPhone 14 hands down in terms of GPU horsepower. The relevant review, though, is performance per watt. This video ( ▶ this video ) from a few months ago explores the power efficiency graphs and that's probably what most phone users really want. Nobody is gaming on their phone until it hits the limits of passive cooling and very few people will need the raw CPU performance for more than a second per page load. I don't even know what intensive single core benchmarks are even good for in real life, maybe Javascript if you're somehow running the JS VM at 100% for minutes straight? That doesn't sound like something I'd want my phone to do! Qualcom's advancements in speed and longevity have been incremental, sometimes even decremental, for years now. Mediatek, previously the chipset for every 100 dollar Chinese phone, keeps closing while Qualcom desperately tries to squeeze just a little more juice out of their cores. Apple's progress is also slowing down, but not nearly as much as their most important competitor's. It's a shame, really. Hopefully Google and Microsoft will develop their own chips for real in the future because you can't just wait for Qualcom anymore. Microsoft REALLY wants a good M1/2 competitor but the other chips in the ARM s
in “Apple’s 2½ year old iPhone 12 is 6% faster than the new Samsung Galaxy S23 Ultra” · View on HN
No, but there is evidence that Geekbench doesn't do a good job of measuring sustained performance. This video from Geekerwan takes more of a sustained performance / thermal throttling perspective. ▶ this video
in “Apple’s 2½ year old iPhone 12 is 6% faster than the new Samsung Galaxy S23 Ultra” · View on HN
Geekerwan on YouTube tests the performance and effiency of different cores at different frequencies. This is one good example where the A55 and A510 (successors to the A53) are graphed at around 15:40: ▶ this video
in “ARM’s Cortex A53: Tiny but Important” · View on HN
>It's so hard to imagine so many years & a microarchitecture remaining so unchanging. It's a shame, because it was the best design from ARM; they're now focusing on Cortex-A7x and Cortex-X, which aren't anywhere as power efficient[0]. Meanwhile, their revised Cortex-A57 has been surpassed in performance/power/area by several RISC-V microarchitectures, such as SiFive's U74[1], used in the VisionFive2 and Star64, or even the open source XuanTie C910[2][3]. 0. ▶ this video 1. [link] 2. [link] 3. [link]
in “ARM’s Cortex A53: Tiny but Important” · View on HN
It is still the case. This team went deep into benchmarking mobile chipset power efficiency, Exynos does not do well. ▶ this video
in “Google: Turn off VoLTE, Wi-Fi calling: severe Exynos modem vulnerabilities” · View on HN
>I don't think RISC-V rivals you in any market except micro-controllers. I used to think like this. Then I saw the sad state of mobile SoCs[0], and learned that Cortex-A55 is pretty much still the peak, with full awareness of U74[1] and U74-MC[2], which is considerably better (less area, faster, and more efficient to boot). Considering that P650[3] has existed for a while competing with ARM's performance-focused cores, and X280[4], basically an U74 with Vector, also exists today, I simply do not see a 2023 without a pack of new SoC based on these cores (particularly, I expect to see X280 pop up everywhere), and phones based on these SoCs. Android was already demonstrated on RISC-V years ago, and a lot of effort is being put in polishing and upstreaming this work. That's how immediate we're talking about. As for the ARM story we're discussing, it's clear to me: Softbank wants to sell ARM ASAP, and that's about the only thing they care about. They thought that doing this statement is helping them, but I am not sure it is. 0. ▶ this video 1. [link] 2. [link] 3. [link] 4. [link]
in “Arm execs: We respect RISC-V but it's not a rival in the datacenter” · View on HN